{"id":"f5ebc90e-4770-4003-a823-8eef0418bde8","arxiv_id":"2608.00904","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Compact radio sources in the centers of X-ray-selected galaxy clusters can suppress the measured Sunyaev-Zel'dovich signal, causing roughly 4.5% of clusters to be missed by microwave surveys.","lead":"Astronomers used a 90 GHz camera on the Green Bank Telescope to search for bright radio spots in 138 galaxy clusters selected by their X-ray glow. They found such spots are rare, but when they sit near a cluster's center they can hide the cluster's faint microwave signal, explaining why some clusters are missing from all-sky surveys.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Abstract's 4.5% 'missed clusters' overstates the analysis: the paper measures y-suppression, not actual missed detections, and the conclusions contradict the abstract (3% vs 4.5%).","rationale":"The paper presents a valuable observational survey and a careful source-extraction analysis. The detection of 11 compact sources in 138 X-ray-selected clusters, with all having radio counterparts, is a useful contribution. However, the headline statement in the abstract, that these sources 'would result in 4.5% of clusters being missed by tSZ surveys', is not directly supported by the analysis. The analysis computes a suppression proxy (fraction of clusters with y0/ytrue < 0.75) and only labels these as 'potentially missed' in Section 5. The leap from this proxy to an actual missed fraction requires a detection simulation that includes the cluster's intrinsic signal-to-noise and the survey's detection threshold. No such simulation is presented. The internal inconsistency between 4.5% (abstract and Section 5) and 3% (Conclusions) further undermines confidence in the exact number, and the small sample size (11 sources, no uncertainty on the 4.5%) makes the claim fragile. The reader's weakest_assumption identified the response functions as a key uncertainty; that is valid and related, but the more fundamental issue is the conceptual mapping from y-suppression to missed clusters. I agree with the CONDITIONAL verdict: the observational results are solid, but the headline claim should be tempered or augmented with a detection simulation. Thus I leave the verdict unchanged, with the condition that the authors either perform such a simulation or revise the abstract to match the 'potentially missed' language and clarify which of the two percentages is correct.","tokens_in":16079,"tokens_out":7493,"duration_ms":79112,"concrete_test":"Forward-model the actual detection: inject the 11 measured sources (flux, position, spectral index) into ACT DR6 maps at the positions of the 138 M2-eFEDS clusters, run the official DR6 cluster detection pipeline (or a faithful public implementation) on both source-free and source-injected maps, and count how many clusters are detected in each case. The difference is the true missed fraction. If it differs from 4.5% (or the 3% in the Conclusions), the abstract's claim is not supported. Also recompute the y-suppression using the DR6 matched-filter response to verify which of the two printed numbers is correct.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in the abstract ('4.5% of clusters being missed') is not directly computed. Section 5 defines a proxy: the fraction of clusters whose measured Compton-y is suppressed by >25% relative to the source-free value, and calls these 'potentially missed'. There is no simulation of the ACT DR5/DR6 matched-filter detection pipeline to show that a 25% suppression actually causes a cluster to fall below the detection threshold. A cluster with a high true y can lose 25% and still be detected; a cluster near the threshold could be missed with only 20% suppression. The abstract drops the qualifier 'potentially', turning a proxy into a definite missed fraction. In addition, the paper is internally inconsistent: the abstract and Section 5 state 4.5% (Section 5) while the Conclusions state 3% for the same quantity, with no explanation. The calculation also rests on response functions N(r) and A(α) from Dicker et al. (2021) that are acknowledged to depend on the specific matched filters; if applied to DR6 rather than DR5, the suppression values change. These issues together mean the headline number is not robust.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a blind 90 GHz MUSTANG2 survey of 138 X-ray-selected eFEDS galaxy clusters, of which 96 are new observations, with the goal of quantifying how compact radio sources can suppress the thermal Sunyaev-Zel'dovich (tSZ) signal and thereby cause clusters to be missed in surveys such as ACT DR5/DR6. The survey reaches a central 5-sigma detection limit near 1 mJy, detects 11 discrete sources (all with radio counterparts), and compares their fluxes and radial distribution with the tSZ-selected M2-ACT sample and with the Websky simulation. Using the D2021 response functions (Equation 1) to convert source flux, projected radius, and spectral index into a change in the measured central Compton-y, the paper reports that 4.5% of the X-ray-selected clusters have measured Compton-y more than 25% below the true value, versus 1.5% for the tSZ-selected and Websky samples, and frames this as a cluster-missed fraction for tSZ surveys.","tokens_in":16404,"tokens_out":5107,"duration_ms":57181,"significance":"If the central claim holds, the paper provides a useful empirical constraint on a known systematic in cluster cosmology: compact-source contamination of tSZ signals. The observational design is a strength: the survey is blind, uses signal-free noise maps and explicit false-positive control, and the source SED matching is careful. The comparison between an X-ray-selected sample and a tSZ-selected sample is well motivated and helps separate selection effects. However, the headline 4.5% figure is currently a proxy for missed clusters rather than a measured detection-completeness loss, it has no quoted uncertainty, and it is inconsistent with the 3% figure in the Conclusions. The result is therefore better viewed as an indication of a few-percent-level effect that needs further qualification and validation before it can support the strong abstract wording.","major_comments":[{"comment":"The headline number is internally inconsistent. The abstract and Section 5 state that 4.5% of the X-ray-selected sample have a measured tSZ signal more than 25% below the true value, while the Conclusions state that the extra central sources 'would result in 3% of clusters having measured tSZ signals lower than their true value by 25%.' This is the same quantity presented as 4.5% and 3% with no explanation. The paper must reconcile these numbers and state explicitly which sample and definition each percentage refers to.","section":"Abstract; Section 5; Section 6"},{"comment":"The leap from y0/ytrue < 0.75 to 'clusters being missed' is not supported. Section 5 defines the 25% suppression threshold and calls those clusters 'potentially missed'; the abstract drops the qualifier and converts this proxy into a definite missed fraction. A cluster with a high true y can lose 25% and still be detected, while a cluster near the detection threshold could be lost with less suppression. No end-to-end simulation of the ACT DR5/DR6 matched-filter detection pipeline is provided. At minimum the abstract must say 'potentially missed' or 'would have measured signals suppressed by >25%,' and ideally the paper should estimate the actual completeness loss using the survey selection function.","section":"Abstract; Section 5"},{"comment":"The calculation relies on the response functions N(r) and A(alpha) taken from Dicker et al. (2021) for the ACT DR5 matched filter. Figure 1 itself acknowledges that 'the exact shape depends on the matched filters used by a given tSZ survey,' yet the paper applies these functions to the new sample and then compares with the DR6-based estimate from ACT DES HSC Collaboration et al. (2025). No sensitivity analysis is given for plausible changes in N(r) or A(alpha), or for a DR6-specific response. Since the null at 104 arcsec and the sign of the response determine the central result, this is a load-bearing systematic that needs to be quantified or justified.","section":"Equation (1); Figure 1; Section 5"},{"comment":"The key comparison of 4.5% (X-ray-selected) versus 1.5% (tSZ-selected and Websky) is based on only 11 detected sources, and the M2-eFEDS sample includes 42 clusters that also appear in the M2-ACT sample, so the two observational samples are not independent. No statistical uncertainty is quoted for the percentages. With 138 clusters, each source corresponds to roughly 0.7%, so the internal 4.5% vs 3% discrepancy and the cross-sample 4.5% vs 1.5% difference may be marginal. The paper should quote Poisson/bootstrap uncertainties and account for the sample overlap when making the comparison.","section":"Table 2; Section 4.1; Section 5"}],"minor_comments":[{"comment":"The number of eFEDS clusters missing from DR5 is given as 96 in the abstract and Section 2.1, but Section 6 says '27 of the 98 clusters seen in eFEDS but not DR5'. Please make the count consistent.","section":"Section 2.1; Section 6"},{"comment":"There is a typo: 'with with the same mapmaker settings' should read 'with the same mapmaker settings'. Also 'medium noise' should presumably be 'median noise'.","section":"Section 2.2"},{"comment":"The factor w appears in Equation (1) but is not defined in the text. Please define all symbols explicitly, including its value and role.","section":"Equation (1)"},{"comment":"The figure caption states that solid and dashed parts of the response line represent regions where the source reduces or increases the measured Compton-y, but it would be clearer to spell this out directly in the caption rather than only in the text.","section":"Figure 3"},{"comment":"The table lists tSZ masses of 0.00 or 'neg Y' for many clusters. The text explains that negative Compton-y values are set to zero, but it would be helpful to clarify how these entries are treated in the mass comparisons and whether they enter the analysis at all.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"This is a solid observational paper in scope for the journal. The main problem is the mismatch between the strong abstract claim ('4.5% of clusters being missed') and the actual analysis (a >25% y-suppression proxy), together with the unexplained 4.5% vs 3% inconsistency. Both issues are fixable with rewording and an explicit sensitivity/uncertainty analysis. No concerns about attribution or novelty beyond what is stated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: the MUSTANG2 survey of 138 eFEDS clusters (96 new) is exactly the kind of systematic check tSZ cosmology needs. The finding that compact sources are more centrally concentrated in X-ray-selected than tSZ-selected clusters is new and plausible, and the paper's claim that this can suppress measured Compton-y for a few percent of clusters is worth taking seriously. But the headline \"4.5% missed\" overstates what is actually measured, and the paper contradicts itself between abstract and conclusions.\n\nWhat is good: the data reduction is careful. Signal-free noise maps, false-positive control via SNR cuts in inner/outer regions, SED fits with radio counterparts. The survey reaches ~1 mJy at 5 sigma in cluster centers. The comparison to their earlier M2-ACT sample and to Websky is honest; the radial profile of source flux is the most interesting result. They do not oversell the radio spectral index data.\n\nSoft spots: the 4.5% figure is derived from 11 sources, and it is the fraction of clusters whose measured y is suppressed by more than 25% relative to the source-free value, not the fraction actually missed by an ACT-like detection pipeline. A 25% suppression is not a detection threshold; it is a proxy. The abstract drops the word \"potentially\" that appears in Section 5. And the conclusions state 3% for the same quantity, with no explanation. That is an internal inconsistency the authors should fix.\n\nAlso, the conversion relies on the D2021 matched-filter response functions N(r) and A(alpha), which the paper itself notes depend on the specific filter. They use ACT DR5 filters; applying to DR6 might change the number. They cite DR6's 2-3% estimate as consistent, but that comparison is loose.\n\nOverall: the qualitative claim — that X-ray-selected clusters have more central contamination that can bias tSZ detection — is likely right. The quantitative 4.5% is not robust as presented. This deserves a referee. I would suggest the authors either present the y-suppression fraction as a function of threshold or run a proper matched-filter injection simulation to convert suppression into actual missed fraction. Also fix the 3% vs 4.5% discrepancy.\n\nWho it is for: anyone working on cluster cosmology with ACT/SPT/SO, or on radio source contamination. I would bring it to a reading group.\n\nRecommendation: send to peer review, but with the expectation of moderate revisions.","headline":"A careful new 90 GHz survey of X-ray-selected clusters that likely finds a real contamination effect, but the headline 4.5% missed-cluster figure is a proxy, not a measured fraction, and the paper contradicts itself between abstract and conclusions.","tokens_in":16880,"tokens_out":2437,"would_cite":true,"duration_ms":24841,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Compact radio sources near cluster centers can push X-ray-selected clusters out of tSZ surveys, and this paper measures the effect at 4.5%.","keywords":["galaxy clusters","Sunyaev-Zeldovich effect","X-ray surveys","compact radio sources","tSZ survey completeness","matched filter","MUSTANG2","eFEDS"],"falsifier":"Re-run the ACT DR5 cluster-detection pipeline on the same 138 eFEDS clusters after subtracting the 11 detected compact sources from the maps; if the number of new detections is not consistent with the predicted 4.5% fraction (about 6 clusters), the matched-filter response model underlying the estimate is wrong.","tokens_in":16050,"feed_emoji":"📡","tokens_out":12586,"duration_ms":124271,"temperature":0.7,"pith_summary":"This paper investigates why some galaxy clusters that should be detectable by millimeter-wave tSZ surveys are missing from tSZ catalogs. Using high-resolution 90 GHz observations of 138 X-ray-selected clusters from eFEDS, it finds 11 compact radio sources, distributed closer to cluster centers than sources in a comparable tSZ-selected sample. When each source is run through the matched-filter response of the ACT DR5 survey (Equation 1), 4.5% of the X-ray-selected clusters have their measured Compton-y suppressed by more than 25%, enough to push them below the detection threshold. The same calculation gives 1.5% for tSZ-selected clusters and the Websky simulation. The paper concludes that compact source contamination creates a small but non-random incompleteness in tSZ cluster surveys, more pronounced for X-ray-selected systems.","feed_headline":"4.5% of X-ray clusters lost to radio sources in tSZ surveys","feed_subtitle":"Centrally placed 90 GHz radio sources suppress the Sunyaev-Zel'dovich signal, causing tSZ surveys to miss clusters.","key_machinery":"The load-bearing tool is Equation 1, $\\Delta \\tilde{y}_0 = I \\, \\delta\\tilde{y}_0 \\, N(r) \\, A(\\alpha)$, which converts a point source's 90 GHz flux $I$, its projected radius $r$ from the cluster center, and its spectral index $\\alpha$ into the resulting change in the ACT DR5 central Compton-$y$. The functions $N(r)$ and $A(\\alpha)$ (Figure 1, from D2021) encode how the survey's matched filter responds to a point source; $N(r)$ has a null at 104 arcseconds, so sources inside this radius suppress the measured tSZ signal while sources outside add to it. This sign change is what makes centrally located sources dangerous for cluster detection and determines the 4.5% missed fraction.","core_discovery":"Central discovery: A blind 90 GHz survey of 138 X-ray-selected eFEDS clusters with MUSTANG2 found 11 compact sources, all with radio counterparts. Crucially, the sources in X-ray-selected clusters are concentrated near cluster centers, where their positive flux cancels the negative tSZ decrement. Applying the ACT DR5 matched-filter response functions (Equation 1) to each detected source predicts that 4.5% of the X-ray-selected sample has a measured Compton-y more than 25% below its true value—enough to miss these clusters—versus 1.5% for tSZ-selected clusters and the Websky simulation. The paper also shows the X-ray masses that placed these clusters above ACT detection limits are biased high","pith_inferences":["The true missed fraction is likely a lower bound: the survey reaches 5σ at 1 mJy, and fainter sources below this threshold will also suppress $y$; deeper 90 GHz imaging of the same fields would test this.","The filter-specific $N(r)$ and $A(\\alpha)$ mean the 4.5% number applies to ACT DR5-like pipelines; other tSZ surveys with different matched filters need their own high-resolution source surveys to quantify equivalent incompleteness.","If central radio sources preferentially trace active galactic nuclei, then the missed clusters may be biased toward systems with recent feedback or merging activity, which could affect scatter in scaling relations."],"forward_implications":["tSZ cluster surveys like ACT DR5 have a source-induced incompleteness of order 4.5% for X-ray-selected clusters and 1.5% for tSZ-selected clusters, which must be folded into cluster count cosmology.","The missed clusters are not uniformly distributed: they preferentially have compact radio sources near their centers, so any cosmological analysis using tSZ cluster samples carries a selection bias tied to radio-loud active galactic nuclei.","The X-ray mass estimates used to build the eFEDS sample are biased high by a factor of 1.9 compared to weak-lensing masses, meaning X-ray-selected samples contain many lower-mass clusters than their X-ray masses suggest.","The source distributions in X-ray- and tSZ-selected clusters differ: X-ray clusters have more central flux, tSZ clusters have more flux at radii >104 arcsec where sources boost the measured signal; this explains why the two selection methods see different contamination patterns.","The 4.5% estimate is consistent with the 2–3% missed-cluster estimate from recent ACT DR6 analyses, suggesting the effect persists even with deeper data."],"supporting_citations":[{"why":"Supplies the matched-filter response functions N(r) and A(α) used in Equation 1 to convert source flux, radius, and spectral index into a change in measured Compton-y.","marker":"Dicker et al. 2021"},{"why":"Provides the tSZ-selected M2-ACT sample of 248 clusters and their source detections, the direct comparison sample for source prevalence and radial distribution.","marker":"Dicker et al. 2024"},{"why":"Provides the ACT DR5 cluster catalog and survey limits that define which eFEDS clusters should have been detected.","marker":"Hilton et al. 2021"},{"why":"Estimates 2–3% of clusters are missed by compact sources in DR6; this paper's 4.5% is compared to that value.","marker":"ACT DES HSC Collaboration et al. 2025"},{"why":"Supplies the eFEDS X-ray cluster catalog and the X-ray mass and redshift estimates used to build the M2-eFEDS sample.","marker":"Liu et al. 2022"},{"why":"Provides the Websky simulation used as a model comparison for source counts and radial distribution.","marker":"Stein et al. 2020"},{"why":"Supplies the radio source population in the Websky simulation, used to compute the simulated contamination.","marker":"Li et al. 2022"},{"why":"Provides weak-lensing masses used to show that X-ray masses are biased high and to compute tSZ masses from forced photometry.","marker":"Chiu et al. 2022"}],"fun_headline_variants":["Centrally placed radio sources hide 4.5% of galaxy clusters","X-ray clusters' radio cores mask tSZ signal, missing 4.5%","MUSTANG2 finds 11 radio sources that suppress tSZ in 138 clusters","Radio sources near cluster centers cancel tSZ, losing 4.5%","eFEDS clusters above tSZ limits: many missed due to radio sources"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The 4.5% missed-cluster fraction depends on the accuracy of the matched-filter response functions N(r) and A(α) from Dicker et al. (2021) in representing the actual ACT DR5 pipeline; if those curves are wrong for real data (for instance, if the filter response changes with angular resolution or source extension), the missed fraction changes.","fun_headline_variants_meta":{"raw":{"variants":["Centrally placed radio sources hide 4.5% of galaxy clusters","X-ray clusters' radio cores mask tSZ signal, missing 4.5%","MUSTANG2 finds 11 radio sources that suppress tSZ in 138 clusters","Radio sources near cluster centers cancel tSZ, losing 4.5%","eFEDS clusters above tSZ limits: many missed due to radio sources"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000878,"raw_usage":{"total_tokens":3684,"prompt_tokens":844,"completion_tokens":2840,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":2736}},"tokens_in":588,"tokens_out":2840,"duration_ms":23211,"temperature":1.0,"reasoning_tokens":2736,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T00:41:40.491280+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the ACT DR5 cluster-detection pipeline on the same 138 eFEDS clusters after subtracting the 11 detected compact sources from the maps; if the number of new detections is not consistent with the predicted 4.5% fraction (about 6 clusters), the matched-filter response model underlying the estimate is wrong.","supporting_citations":[],"review_version":1}